MAINT deprecated 'spectral' in favor of 'nipy_spectral'

[NelleV]

This raises a deprecation warning for the deprecated "spectral" colormap.
I think I covered all the possible calls to "spectral".

closes #7315

[NelleV]

@Carreau Do you want to review? :)

[Carreau]

@Carreau Do you want to review? :)

Yes :-) The name vas changed in 2012, so I would not be shy as marking as deprecated since at least matplotlib 1.5.

You might want to warn on spectral_r as well (constructed in lib/matplotlib/cm.py)

For post 2.0 / Subsequent PRs:

• It might be hard, but add a warning filter that turns use of spectral during testing (and doc building ?) into an error.
• Have keys(), iterkeys(), items() not list spectral and spectral_r

An easy implementation would be to actually remove the data find the dict, and if name==spectral, return super().__getattra__('nipy_spectral_'). But that can be done later.

[Carreau]

doc/users/image_tutorial.rst uses spectral...

[NelleV]

Yeah, I forgot to look at the documentation and "fix" it…

[Carreau]

Sorry I'm slow to review I'm in the train, tethering and the car is shaking quite a bit.

[NelleV]

Updated it to deprecate spectral_r as well.

[efiring]

Looks OK to me.

[Carreau]

do you want to make the message dynamic for spectral_r as well ? I can see pros and cons to both.

[NelleV]

I was going to be lazy, and not do it, but… maybe I should.

[Carreau]

Note that line 88 of this file will likely trigger the deprecation warning if I'm not wrong, as it loop through all the colormap to build the reversed ones. Not sure what we can do about the though. Just pointing it out.

[NelleV]

argh… you are right.
I'll fix that.

[Carreau]

Updated it to deprecate spectral_r as well.

+1. Thanks @NelleV !

[NelleV]

I'm now catching warnings when creating the reverse colormaps to avoid raising the deprecation warning when cm.py is imported (ie, when anyone uses matplotlib).

[tacaswell]

Do we want to also deprecate spectral function in pyplot as well?

[NelleV]

Yes, we should… but how to do this with the function being generated?

[tacaswell]

https://github.com/matplotlib/matplotlib/blob/v2.x/boilerplate.py#L333 split the list into two (deprecated and not) and make a new template which includes the deprecation warning.

[NelleV]

Got it. I'll do that.

[NelleV]

Should I run boilerplate and commit the function as well or should we wait for the release?

[efiring]

Yes, it's best to keep pyplot in sync with everything else.

[QuLogic]

Wow, pyplot.py seems to be pretty outdated; will have to be careful when backporting.

[tacaswell]

Should probably just re-run boilerplate after doing the backport.

[efiring]

backported as f87100a (including the pyplot.py regenerated by boilerplate.py)